Soil Type Prediction using Soil Data Grevena

1. Domain

Soil Types — Predicting soil type based on chemical and physical properties.

2. Dataset

• Source: Kaggle - Soil Data Grevena
• File: SOIL DATA GR.xlsx
• Attributes: The dataset contains various chemical and physical properties of soil samples, such as pH, Electrical Conductivity, and the percentage of Sand, Silt, and Clay.
• Target Variable: The Soil Type column was not present in the original dataset. For the purpose of this analysis, a synthetic Soil Type column was generated using the KMeans clustering algorithm. This allows us to demonstrate a complete classification workflow.

3. Problem Statement

Given the chemical and physical properties of soil, predict the soil type. This project tackles this as a multi-class classification problem.

4. Methodology

a. Data Loading & Preprocessing

• Data Loading: The data was loaded from the SOIL DATA GR.xlsx Excel file using the pandas library.
• Handling Missing Values: Rows with any missing values were dropped to ensure data quality.
• Column Name Cleaning: Leading and trailing whitespaces were stripped from all column names to prevent potential errors during data manipulation.
• Synthetic Target Variable: As the original dataset lacks a target variable for soil type, the KMeans clustering algorithm was employed to create a synthetic Soil Type column. The data was grouped into three clusters, which were then used as the target labels for the classification task.
• Label Encoding: The categorical Soil Type labels ('0', '1', '2') were converted into a numerical format using LabelEncoder from scikit-learn.
• Feature Scaling: All numerical features were scaled using StandardScaler to standardize the data and improve the performance of the machine learning models.
• Outlier Detection: Outliers were detected using the z-score method. Data points with a z-score greater than 3 were identified as outliers and removed from the dataset.

b. Exploratory Data Analysis (EDA)

• Distribution of Soil Types: A countplot was generated to visualize the distribution of the synthetically generated soil types.
• Feature Correlation: A correlation heatmap was created to explore the relationships between the different soil properties.
• Pairplot: A pairplot was used to visualize the relationships between pairs of features, with the points colored by the synthetic soil type.

c. Model Building

• Classification:
  • Two classification algorithms were applied:
    • Random Forest Classifier: An ensemble learning method that operates by constructing a multitude of decision trees at training time and outputting the class that is the mode of the classes of the individual trees.
    • Logistic Regression: A linear model for classification rather than regression.
  • The models were evaluated using the following metrics:
    • Accuracy: The proportion of correctly classified samples.
    • Classification Report: A detailed report showing the precision, recall, and F1-score for each class.
    • Confusion Matrix: A table used to describe the performance of a classification model on a set of test data for which the true values are known.
• Clustering:
  • KMeans Clustering: The KMeans algorithm was applied to the data to group it into clusters based on their features. The results were presented in a cross-tabulation against the synthetic soil types.

d. Results

• Random Forest Classifier: Achieved an accuracy of 1.0. While this is a perfect score, it is highly likely that the model is overfitting the data. This is a common issue with tree-based models, especially when the target variable is generated from the same features used for training.
• Logistic Regression Classifier: Achieved an accuracy of approximately 0.96, which is a very good result, but may also be influenced by the synthetic nature of the target variable.
• KMeans Clustering: The algorithm successfully grouped the data into three distinct clusters, which were then used as the basis for the classification task.

5. How to Run

1. Download the Dataset: Download the SOIL DATA GR.xlsx file from the Kaggle dataset page and place it in the same directory as the script.
2. Install Dependencies: Make sure you have Python installed, and then install the required libraries using pip:

   pip install pandas numpy matplotlib seaborn scikit-learn openpyxl

3. Run the Script: Execute the following command in your terminal:

   python soil_type_prediction.py

6. Conclusion

This project demonstrates a complete predictive analysis workflow, from data loading and preprocessing to model building and evaluation. While the use of a synthetically generated target variable limits the real-world applicability of the classification results, the project serves as a comprehensive example of how to approach a data science problem when faced with an unsupervised learning task that can be framed as a supervised one. The high accuracy scores, while promising, should be interpreted with caution due to the nature of the target variable's creation.